Railway tunnel structure point cloud decoupling separation method

Abstract

The application discloses a railway tunnel structure point cloud decoupling separation method and relates to the technical field of railway tunnel point cloud processing. The method obtains single-layer tunnel point cloud through voxelization and downsampling; applies a CSF algorithm to separate track point cloud; divides the remaining point cloud into a left sidewall block, a right sidewall block and a vault block; calculates the main plane of each block through PCA and rotates to a parallel horizontal reference surface; applies the CSF algorithm again to separate the sidewall structure and the attached object point cloud; restores and splices the coordinates to obtain a fine separation result. The method creatively rotates the vertical wall surface to a virtual horizontal surface through PCA rotation, solves the problem that the CSF algorithm is invalid in a vertical scene, combines a double-threshold design, realizes a separation success rate of 90%+ in a complex scene, shortens the single-kilometer processing time to 20 minutes, and improves the railway tunnel point cloud processing efficiency and accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of railway tunnel point cloud processing, and particularly relates to a railway tunnel structure point cloud decoupling separation method. BACKGROUND

[0002] In the field of railway tunnel point cloud processing, in view of the separation requirement of the main structure and the attached object, the existing technical system has a technical bottleneck that is difficult to break through, which is specifically manifested in the following three core limitations:

[0003] The traditional filtering segmentation method (such as statistical filtering, region growing algorithm, etc.) excessively relies on the experience parameters set by artificial, and in the processing of complex tunnel scenes, it is difficult to solve the problem of adhesion point cloud segmentation of the track and the wall surface, and it is easy to produce missegmentation for the curved surface structure (such as the arch top arc area of the tunnel), which leads to the fuzzy boundary of the main structure and the non-structure object, and seriously affects the subsequent detection and modeling accuracy.

[0004] Although the semantic segmentation technology based on deep learning shows certain potential in some scenes, this technology highly depends on the support of large-scale labeled data, and there are a large number of small attached objects (such as bolts, cables, embedded parts, etc.) in the railway tunnel scene, which are difficult to extract features, and are prone to miss detection; at the same time, the generalization ability of this technology for cross-scenarios (such as different geological conditions and different tunnel types) is weak, and it is difficult to adapt to the diversified application requirements in actual engineering.

[0005] The conventional cloth simulation filtering (CSF) algorithm is based on the design of the gravity field physical model, and there is a natural conflict between the filtering direction and the vertical wall of the tunnel, which leads to the complete failure in processing the side wall attached object (such as the catenary, pipeline support, etc.), and cannot realize the accurate separation of the main structure and the non-structure object on the vertical surface, which greatly limits the engineering practicability of the algorithm in the tunnel point cloud processing.

[0006] Therefore, how to break through the limitations of the existing technology in parameter dependence, data requirement and scene adaptability, and realize the efficient and accurate separation of the main structure and the attached object in the railway tunnel point cloud, has become a key problem for promoting the development of intelligent detection and modeling technology of transportation infrastructure. SUMMARY

[0007] The purpose of the present application is to provide a railway tunnel structure point cloud decoupling separation method to solve the above problems existing in the prior art, realize the automatic segmentation of complex elements such as tunnel wall, arch top, track and pipeline, and improve the separation precision and efficiency.

[0008] In order to achieve the above purpose, the present application is realized by the following technical scheme: a railway tunnel structure point cloud decoupling separation method, comprising the following steps:

[0009] Step S1, voxelization downsampling processing is performed on the original railway tunnel point cloud to obtain a single-layer tunnel point cloud model;

[0010] Step S2, a cloth simulation filtering algorithm is applied to separate the point cloud of the tunnel bottom track area;

[0011] Step S3, the remaining tunnel point cloud is divided into three independent structural blocks, i.e., a left sidewall block, a right sidewall block and a vault block;

[0012] Step S4, directional adjustment is performed on each structural block, the best fitting plane of each point cloud block is calculated through principal component analysis, a rotation matrix is calculated based on the plane normal vector, and the main plane of each point cloud block is rotated to be parallel to the horizontal reference surface;

[0013] Step S5, the cloth simulation filtering algorithm is applied again to each point cloud block after rotation transformation to separate the tunnel sidewall structure point cloud and the non-structure point cloud attached thereto;

[0014] Step S6, the inverse matrix of the corresponding rotation matrix is applied to each tunnel sidewall point cloud block extracted to restore it to the original coordinate system;

[0015] Step S7, the three tunnel structure point cloud blocks after inverse transformation are spliced to form a complete tunnel main structure point cloud, and uniformization processing is performed to obtain a fine separated tunnel structure surface point cloud.

[0016] Further improvement of the technical scheme of the present application, in the voxelization downsampling processing, the voxel size is set to be slightly higher than the nominal accuracy of the point cloud collection instrument.

[0017] Further improvement of the technical scheme of the present application, in step S2, when the cloth simulation filtering algorithm is applied to separate the point cloud of the tunnel bottom track area, the grid size is set to be ≤0.1 meters, the separation threshold is 0.2 meters to 0.5 meters, and the separation threshold is strictly higher than the actual pavement height of the track.

[0018] Further improvement of the technical scheme of the present application, in step S3, dividing the remaining tunnel point cloud into three independent structural blocks includes the following sub-steps:

[0019] Step S31, a copy of the original point cloud is created, the copy point cloud is vertically cut along the tunnel axis direction to divide it into left and right parts, which are respectively used as the point cloud of the left sidewall block and the right sidewall block;

[0020] Step S32, along the tunnel axis direction, the original point cloud that has not been divided is cut perpendicularly to the division line in step S31 to divide it into upper and lower parts, and the upper part point cloud is reserved as the point cloud of the vault block.

[0021] Further, as an improvement of the technical scheme of the present application, the directional adjustment of each structure block in step S4 further comprises: calculating the minimum bounding box center of each point cloud block, and taking the minimum bounding box center as the rotation center of the rotation transformation.

[0022] Further, as an improvement of the technical scheme of the present application, in step S5, when the cloth simulation filtering algorithm is applied to each point cloud block after the rotation transformation, the grid size is set to be less than or equal to 0.1 meter, and the separation threshold is set to be less than or equal to 0.1 meter.

[0023] Further, as an improvement of the technical scheme of the present application, in step S5, the non-structure point cloud comprises a contact network, a cable pipeline and a distribution box.

[0024] Further, as an improvement of the technical scheme of the present application, in step S6, the homogenization processing is performed by using an initial voxel size.

[0025] The second aspect of the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the railway tunnel structure point cloud decoupling and separation method when executing the computer program.

[0026] The third aspect of the present application provides a computer readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the railway tunnel structure point cloud decoupling and separation method.

[0027] The present application has the following beneficial effects:

[0028] The most core innovation of the present application is to propose a "virtual horizontal plane space reconstruction closed loop processing" mechanism: through PCA rotation transformation, the vertical tunnel wall is dynamically redirected to a virtual horizontal plane, which completely releases the separation potential of the CSF algorithm in the vertical scene; combined with the "voxel precision self-adaptation - double threshold engineering design - inverse transformation geometric reduction" whole process, the three major bottlenecks of the existing technology, i.e., the CSF gravity direction failure, the traditional filtering structure adhesion and the deep learning labeling dependence, are systematically solved. The whole process automation processing compresses the single kilometer tunnel point cloud analysis time from 1 hour to 20 minutes. The separation success rate in complex scenes such as curved tunnels is 90%+, which provides a high robustness technical base for railway intelligent operation and maintenance. BRIEF DESCRIPTION OF DRAWINGS

[0029] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments, made with reference to the accompanying drawings:

[0030] Figure 1 A whole process flowchart of the railway tunnel structure point cloud decoupling and separation method according to some embodiments of the present application is shown.

[0031] Figure 2 Point cloud diagram for experiments according to some embodiments of the application;

[0032] Figure 3 Point cloud diagram for experiments according to some embodiments of the application;

[0033] Figure 4 Point cloud diagram for experiments according to some embodiments of the application;

[0034] Figure 5 Point cloud diagram for experiments according to some embodiments of the application;

[0035] Figure 6 Point cloud diagram for experiments according to some embodiments of the application;

[0036] Figure 7 Point cloud diagram for experiments according to some embodiments of the application;

[0037] Figure 8 Point cloud diagram for experiments according to some embodiments of the application;

[0038] Figure 9 Point cloud diagram for experiments according to some embodiments of the application; DETAILED DESCRIPTION

[0039] The application will be described in greater detail with reference to the drawings and specific embodiments. The schematic embodiments and descriptions of the application are used to explain the application, but are not intended to limit the application.

[0040] It should be noted that all directional directions (such as up, down, left, right, front, back, upper end, lower end, top, bottom, etc.) in the embodiments of the application are only used to explain the relative positional relationship, movement condition, etc. between the components in a certain posture (as shown in the drawings), and if the certain posture changes, the directional directions will also change accordingly.

[0041] In the present application, unless otherwise explicitly specified and limited, the term "connection" should be understood broadly, for example, "connection" can be fixed connection, or detachable connection, or integral; can be mechanical connection, or electrical connection; can be directly connected, or indirectly connected through an intermediate medium; can be internal communication of two elements or interaction relationship between two elements, unless otherwise explicitly limited. For those skilled in the art, the specific meaning of the above-mentioned term in the present application can be understood according to the specific circumstances.

[0042] In addition, the description such as "first", "second" and the like in the present application is only for the purpose of description and cannot be understood as indicating or implying the relative importance of the technical features or implying the number of the technical features indicated. Therefore, the features defined as "first", "second" can be explicitly or implicitly included at least one of the features; in addition, the technical solutions of various embodiments can be combined with each other, but it must be based on the realization of the ordinary skilled in the art, when the combination of technical solutions appears contradictory or cannot be realized, it should be considered that the combination of technical solutions does not exist, also not within the protection scope required by the present application.

[0043] The present application will be further described in detail below with reference to the accompanying drawings.

[0044] Referring to Figure 8 , the first aspect of the present application provides a railway tunnel structure point cloud decoupling separation method, comprising the following steps:

[0045] Step S1, voxelization downsampling processing is performed on the original railway tunnel point cloud to obtain a single-layer tunnel point cloud model;

[0046] Step S2, a cloth simulation filtering algorithm is applied to separate the point cloud of the tunnel bottom track area;

[0047] Step S3, the remaining tunnel point cloud is divided into three independent structure blocks of left side wall block, right side wall block and vault block;

[0048] Step S4, directional adjustment is performed on each structure block, the best fitting plane of each point cloud block is calculated through principal component analysis, and a rotation matrix is calculated based on the plane normal vector, so that the main plane of each point cloud block is rotated to be parallel to the horizontal reference surface;

[0049] Step S5, the cloth simulation filtering algorithm is applied again to each point cloud block after rotation transformation to separate the tunnel side wall structure point cloud and the non-structure object point cloud attached thereto;

[0050] Step S6, the inverse matrix of the corresponding rotation matrix is applied to each tunnel side wall point cloud block extracted to restore it to the original coordinate system;

[0051] Step S7, the three tunnel structure point cloud blocks after inverse transformation are spliced to form a complete tunnel main structure point cloud, and uniformization processing is performed to obtain a fine separation tunnel structure surface point cloud.

[0052] Referring to Figures 1 to 7 , it should be noted that the railway tunnel structure point cloud decoupling separation method of the present application specifically comprises the following steps:

[0053] Step 1: Voxelization down-sampling. Set the voxel size according to the point cloud collection instrument precision, and perform voxelization processing on the original point cloud to obtain a single-layer tunnel point cloud model. Control condition: The voxel size of voxelization processing should be set to be slightly higher than the nominal precision of the point cloud collection instrument (for example, if the instrument precision is ±5mm, the voxel size can be set to 10-15mm). Purpose: This setting can effectively avoid potential point cloud layering abnormalities caused by the dispersion of the original point cloud measurement precision.

[0054] Step 2: Track point cloud separation. Apply the CSF algorithm to separate the point cloud of the tunnel bottom track area. Control condition: When applying the CSF algorithm to separate the track point cloud, two key parameters need to be set: grid size ≤0.1 meters; the separation threshold is recommended to be in the range of 0.2m to 0.5m, and the set value must be strictly higher than the actual pavement height of the track. The optimal precise parameter value needs to be determined through experimental verification according to the specific point cloud data characteristics and tunnel environment (such as track type, point density).

[0055] Step 3: Tunnel structure block. Divide the remaining tunnel point cloud into three independent structure blocks: left sidewall block, right sidewall block, and vault block. Including A, B two sub-steps:

[0056] A: Point cloud replication and left-right segmentation: create a copy of the original point cloud. Perform vertical cutting on the copy point cloud along the tunnel axis direction (for example, X axis or Z axis, need to be clear according to the actual coordinate system, here it is assumed to be the axis direction in XOZ plane), and divide it into left and right two parts evenly. These two parts are respectively the point cloud of the left sidewall block and the right sidewall block.

[0057] B: Vault point cloud extraction: along the same tunnel axis direction, on the original point cloud which has not been segmented, perform cutting perpendicular to the segmentation line of step A (for example, if step A is longitudinal cutting, this step is transverse cutting), divide it into upper and lower two parts. Keep the upper part of the point cloud as the point cloud of the vault block.

[0058] Step 4: Point cloud block orientation adjustment. Including A, B, C, D four sub-steps:

[0059] A: Main plane calculation: apply PCA (Principal Component Analysis) to the point cloud of the left sidewall block, right sidewall block and vault block respectively, and calculate the best fitting plane (i.e. PCA main plane) of each point cloud block.

[0060] B: Rotation center determination: calculate the minimum bounding box center of each point cloud block, and take it as the rotation center of the subsequent rotation transformation.

[0061] C: Rotation matrix calculation: Based on the principal plane normal vector obtained in step A, calculate the rotation transformation matrix required to rotate the principal plane of each point cloud block to be parallel to the horizontal reference plane (XOY plane).

[0062] D: Point cloud transformation and orientation: Apply the corresponding rotation matrix calculated in step C to perform a spatial rotation transformation on the respective point cloud block. This rotation transformation aims to ensure that the outer surface of each side wall block point cloud (i.e., the side facing the outside of the tunnel) is adjusted downward.

[0063] Step 5: Reapply the CSF algorithm to each point cloud block after rotation transformation (at this time, the tunnel side wall surface is simulated as "ground"). Through this step, the tunnel side wall structure point cloud (as "ground") and the non-structure point cloud attached to it (such as catenary, cable pipeline, distribution box, etc.) are effectively separated. Control conditions: When applying the CSF algorithm to finely separate the side wall attachments, two key parameters need to be set: grid size ≤ 0.1 meters; the separation threshold value is recommended to be set to ≤ 0.1 meters. Parameter explanation: Similar to step 2, the optimal precise parameter value needs to be determined through experimental optimization based on the actual processing effect and point cloud quality (such as side wall smoothness, attachment size).

[0064] Step 6: Coordinate restoration: Apply the inverse matrix of the corresponding rotation matrix to the tunnel side wall point cloud block extracted in step 5 (which is still in the rotated pose of step 4) to restore its position and orientation to the original coordinate system.

[0065] Step 7: Result integration and post-processing: splice the three tunnel structure point cloud blocks obtained after inverse transformation to form a complete tunnel main structure point cloud, and finally apply the initial voxel size for uniformization processing to obtain a finely separated tunnel structure surface point cloud.

[0066] Figure 1 is a full process flowchart of the railway tunnel structure point cloud decoupling separation method according to some embodiments of the present application. This figure schematically shows the complete processing flow from the original railway tunnel point cloud, through voxelization down-sampling, track separation, structure block, PCA spatial reorientation, attachment separation, coordinate inverse transformation to final splicing and uniformization, clearly presenting the correlation between each step and the transformation process of the processing object, with the parameter settings (grid size, threshold range) of the CSF algorithm in different steps and the processing path of each structure block (left side wall block, right side wall block, vault block) being explicitly marked.

[0067] Figure 2 is a schematic diagram of an experimental point cloud according to some embodiments of the present application. The figure shows the local details of the original railway tunnel point cloud data, including tunnel walls, tracks, and cables attached to the walls, etc. elements, intuitively presenting the original state and complex scene characteristics of the point cloud to be processed, and providing practical scene basis for the necessity of subsequent processing steps.

[0068] Figure 3 is a schematic diagram of separating track point clouds according to some embodiments of the present application. The figure shows the separation effect of the tunnel bottom track point cloud and other structure point clouds after applying the CSF algorithm (grid size ≤ 0.1 m, threshold 0.2-0.5 m), clearly presenting the results of accurately extracting the track area point cloud, and verifying the effectiveness of the track separation strategy in step 2.

[0069] Figure 4 is a schematic diagram of three-structure block separation filtering results according to some embodiments of the present application. The figure shows the separation effect of the main structure point cloud and the attached point cloud after the left wall block, the right wall block, and the vault block are processed by PCA spatial reorientation and the CSF algorithm (grid size ≤ 0.1 m, threshold ≤ 0.1 m), which clearly distinguishes the main part of each structure block and the attached non-structure (such as catenary, cable pipeline), and embodies the processing effect of structure block and directional separation in steps 3 to 5.

[0070] Figures 5 to 7 Figure 5 is a schematic diagram of the overall splicing effect, front view splicing effect, and top view splicing effect of the final splicing result according to some embodiments of the present application. The three figures together show the final form of the complete tunnel main structure point cloud after coordinate inverse transformation, splicing, and uniformization processing, clearly presenting the continuity and integrity of the tunnel wall, vault, and other main structures, and verifying the technical effect of fine separation after the whole process.

[0071] The second aspect of the present application provides a computer device comprising a memory and a processor, wherein the memory stores a code, and the processor is configured to acquire the code and execute the railway tunnel structure point cloud decoupling separation method described above.

[0072] The third aspect of the present application provides a computer readable storage medium storing a computer program, wherein the computer program is executed by a processor to implement the railway tunnel structure point cloud decoupling separation method described above. Embodiments

[0073] Reference Figures 1 to 7 A railway tunnel structure point cloud decoupling separation method comprises:

[0074] Step 1: Voxelization and downsampling. According to the accuracy of the point cloud acquisition instrument, the voxel size is set, and the original point cloud is voxelized to obtain a single-layer tunnel point cloud model.

[0075] Step 2: Track point cloud separation. Apply CSF (cloth simulation filter) algorithm to separate the point cloud of the track area at the bottom of the tunnel.

[0076] Step 3: Tunnel structure segmentation. Divide the remaining tunnel point cloud into three independent structure blocks: left sidewall block, right sidewall block, and vault block.

[0077] Step 4: Point cloud block orientation adjustment.

[0078] Step 5: Apply CSF algorithm again to each point cloud block after rotation transformation (at this time the tunnel sidewall surface is simulated as "ground"). Through this step, effectively separate the tunnel sidewall structure point cloud (as "ground") from the non-structure point cloud attached to it (such as catenary, cable pipeline, distribution box, etc.).

[0079] Step 6: Coordinate restoration: apply the inverse matrix of the corresponding rotation matrix to the tunnel sidewall point cloud block extracted in step 5 (which is still in the rotated pose after step 4), to restore its position and orientation to the original coordinate system.

[0080] Step 7: Result integration and post-processing: splice the three tunnel structure point cloud blocks obtained after inverse transformation to form a complete tunnel main structure point cloud, and finally apply the initial voxel size for uniformization processing to obtain the fine separation of the tunnel structure surface point cloud.

[0081] Refinement steps:

[0082] Step 3 includes A and B sub-steps:

[0083] A: Point cloud duplication and left-right segmentation: create a copy of the original point cloud. Cut the first group of point clouds (i.e. the copy point cloud) vertically along the tunnel axis direction (e.g. X axis or Z axis, need to be clear according to the actual coordinate system, here we assume the axis direction in XOZ plane), and evenly divide it into left and right two parts. These two parts are respectively used as the point cloud of the left sidewall block and the right sidewall block.

[0084] B: Vault point cloud extraction: along the same tunnel axis direction, cut the second group of point clouds (i.e. the original point cloud without segmentation, or other unused parts) perpendicular to the segmentation line of step A (for example, if step A is longitudinal cutting, this step is transverse cutting), divide it into upper and lower two parts. Keep the upper part of the point cloud as the point cloud of the vault block.

[0085] Step 4 includes A, B, C, and D sub-steps:

[0086] A: Principal plane calculation: Apply PCA (Principal Component Analysis) to the point cloud of left sidewall block, right sidewall block and vault block respectively, and calculate the best fitting plane of each point cloud block (i.e. the PCA principal plane).

[0087] B: Rotation center determination: Calculate the minimum bounding box center of each point cloud block, and take it as the rotation center of subsequent rotation transformation.

[0088] C: Rotation matrix calculation: Based on the principal plane normal vector obtained in step A, calculate the rotation transformation matrix required to rotate the principal plane of each point cloud block to be parallel to the horizontal reference plane (XOY plane).

[0089] D: Point cloud transformation and orientation: Apply the corresponding rotation matrix calculated in step C to perform spatial rotation transformation on the respective point cloud block. This rotation transformation aims to ensure that the outer surface of each sidewall block point cloud (i.e. the side facing the outside of the tunnel) is adjusted downward.

[0090] Special processing or control or operating conditions:

[0091] Step 1: Voxelization down-sampling

[0092] Control condition: The voxel size of the voxelization process should be set slightly higher than the nominal accuracy of the point cloud acquisition instrument (for example, if the instrument accuracy is ±5mm, the voxel size can be set to 10-15mm).

[0093] Purpose: This setting can effectively avoid potential point cloud layering anomalies caused by the dispersion of the original point cloud measurement accuracy.

[0094] Step 2: Track point cloud separation

[0095] Control condition: When applying the CSF algorithm to separate the track point cloud, two key parameters need to be set:

[0096] Grid size: This parameter should be ≤ 0.1 meters (m).

[0097] Cloth rigidity / distance threshold: The recommended parameter range is 0.2 m to 0.5 m, and the set value must be strictly higher than the actual track pavement height.

[0098] The optimal precise parameter value needs to be determined experimentally based on the specific point cloud data characteristics and tunnel environment (such as track type, point density).

[0099] Step 5: Control condition: When applying the CSF algorithm to finely separate sidewall attachments, two key parameters need to be set:

[0100] Grid Size: This parameter should be ≤ 0.1 meter (m). Small grid size is beneficial for capturing the details of the attachments.

[0101] Cloth Rigidity / Distance Threshold: The suggested parameter setting is ≤ 0.1 meter (m).

[0102] Parameter Explanation: Similar to Step 2, the optimal precise parameter value should be determined through experimental tuning based on the actual processing effect and point cloud quality (such as the smoothness of the side wall and the size of the attachments).

[0103] In addition, the present application also provides a computer device, which refers to Figure 9 , and the computer device includes a memory and a processor, the memory stores code, and the processor is configured to acquire the code and execute the railway tunnel structure point cloud decoupling separation method described above.

[0104] In some embodiments, the railway tunnel structure point cloud decoupling separation method in the above embodiments can be implemented by a computer device, which includes at least one processor, a communication bus, a memory, and at least one communication interface.

[0105] The processor can be a general central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0106] The communication bus can be used to transmit information between the above components.

[0107] The memory can be read-only memory (ROM) or other type of static storage devices that can store static information and instructions, random access memory (RAM) or other type of dynamic storage device that can store information and instructions, electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disk storage, optical storage (including a compact disc, laser disc, optical disc, digital versatile disc, Blu-ray disc, etc.), magnetic storage device, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and that can be accessed by a computer, but not limited to. The memory can exist independently, and is connected with the processor through a communication bus. The memory can also be integrated with the processor.

[0108] The memory is configured to store program code for implementing the scheme of the present application, and the processor is configured to control the execution. The processor is configured to execute the program code stored in the memory. The program code can include one or more software modules. The railway tunnel structure point cloud decoupling and separation method in the above embodiments can be implemented by one or more software modules in the program code of the processor and the memory.

[0109] The communication interface is configured to communicate with other devices or communication networks using any transceiver-like device, such as an Ethernet, radio access network (RAN), wireless local area network (WLAN), etc.

[0110] In a specific implementation, as an embodiment, the computer device can include a plurality of processors, each of which can be a single-CPU processor or a multi-CPU processor. The processor herein can refer to one or more devices, circuits, and / or processing cores for processing data (e.g., computer program instructions).

[0111] The computer device described above can be a general-purpose computer device or a special-purpose computer device. In a specific implementation, the computer device can be a desktop computer, a laptop computer, a network server, a personal digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. The embodiments of the present application do not limit the type of computer device.

[0112] In addition, the application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to realize the railway tunnel structure point cloud decoupling separation method.

[0113] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application is further described below in combination with actual scenes:

[0114] A railway tunnel structure point cloud decoupling separation method, the specific steps are as follows:

[0115] Step 1: voxelization and downsampling. The nominal accuracy of the point cloud collection instrument is known to be ±5mm, so the voxel size is set to 12mm, the original railway tunnel point cloud is voxelized, redundant data is removed, a single-layer tunnel point cloud model is obtained, and potential point cloud layering abnormal phenomena caused by the discrete measurement accuracy of the original point cloud are effectively avoided.

[0116] Step 2: track point cloud separation. The CSF algorithm is applied to separate the point cloud of the tunnel bottom track area, the grid size is set to 0.1 meters, and the separation threshold is set to 0.3 meters (which is higher than the actual paving surface height of the track). Through the parameter setting, the track point cloud can be accurately separated, laying a foundation for subsequent structure blocking.

[0117] Step 3: tunnel structure blocking.

[0118] A: Create a copy of the original point cloud, and vertically cut the copy point cloud along the tunnel axis direction (assuming the X-axis direction), and uniformly divide it into left and right two parts as the point cloud of the left and right side wall blocks.

[0119] B: Along the X-axis direction, vertically cut the original point cloud which has not been cut along the cutting line (i.e. the Y-axis direction) to divide it into upper and lower two parts, and keep the upper part as the point cloud of the vault block.

[0120] Step 4: point cloud block directional adjustment.

[0121] A: Apply PCA to the point clouds of the left and right side wall blocks and the vault block respectively, and calculate the best fitting plane of each point cloud block.

[0122] B: Calculate the minimum bounding box center of each point cloud block, and take it as the rotation center of the rotation transformation.

[0123] C: Based on the principal plane normal vector obtained in step A, calculate the rotation transformation matrix required to rotate the principal plane of each point cloud block to be parallel to the XOY plane.

[0124] D: Apply the rotation matrix above to each point cloud block for spatial rotation transformation, ensuring that the outer surface of each side wall block point cloud is adjusted downward.

[0125] Step 5: Apply the CSF algorithm again to each point cloud block after rotation transformation, with a grid size of 0.1 meters and a separation threshold of 0.08 meters. Through this step, the tunnel side wall structure point cloud and the non-structure point cloud such as the catenary, cable pipeline, and distribution box attached to it are effectively separated.

[0126] Step 6: Coordinate restoration. Apply the inverse matrix of the corresponding rotation matrix to each tunnel side wall point cloud block extracted in step 5 to restore its position and orientation to the original coordinate system.

[0127] Step 7: Result integration and post-processing. Splice the three tunnel structure point cloud blocks obtained after inverse transformation to form a complete tunnel main structure point cloud, and apply the initial voxel size (12mm) for uniformization processing to obtain a fine separation of the tunnel structure surface point cloud.

[0128] Through the method of this embodiment, efficient and accurate separation of railway tunnel structure point cloud is achieved, with single-kilometer tunnel point cloud analysis time reduced to 20 minutes, and separation success rate reaching more than 90% in complex scenarios such as curved tunnels, significantly improving the efficiency and quality of railway tunnel point cloud processing.

[0129] In summary, compared with the prior art, the present application has the following beneficial effects:

[0130] Breakthrough vertical scene technical bottleneck: creatively proposes a "virtual horizontal plane space reconstruction closed-loop processing" mechanism, dynamically converts the vertical tunnel wall to a "virtual horizontal plane" through PCA main plane rotation, completely solves the side wall attachment separation failure problem caused by the conflict between the gravity direction and the vertical wall surface of the conventional CSF algorithm, and for the first time realizes the effective application of the CSF algorithm in the vertical curved surface scene, filling the technical gap in this field.

[0131] Full-process automation and precision improvement: combined with the complete technical chain of "voxel precision self-adaptation - double-threshold engineering design - inverse transformation geometric restoration", the traditional filtering segmentation method is systematically overcome, which relies on empirical parameters and is prone to structure adhesion and missegmentation, and the semantic segmentation based on deep learning relies on large-scale labeled data and small-sized attachments are prone to missing. In complex scenarios such as curved tunnels, the separation success rate of the main structure and attachments is improved to more than 90%, which is significantly better than the prior art.

[0132] The engineering efficiency is significantly optimized: the automatic processing mechanism of the whole process greatly shortens the point cloud analysis time, the processing time of single-kilometer railway tunnel point cloud is shortened from 1 hour of the prior art to 20 minutes, the efficiency is increased by about 200%, and the actual demand of large-scale and high-efficiency detection of railway engineering is effectively met.

[0133] Support railway intelligent operation and maintenance upgrade: the fine separation result provided by the application provides high-robustness basic data support for intelligent operation and maintenance links such as railway tunnel structure health monitoring, disease identification and three-dimensional modeling, promotes the transformation of traffic infrastructure detection from traditional manual operation to intelligentization and automation, and has important engineering practical value.

[0134] The technical solutions provided by the embodiments of the application are described in detail above, specific examples are applied in this paper to describe the principles and implementation modes of the embodiments of the application, and the above embodiment descriptions are only applicable to help understand the principles of the embodiments of the application; meanwhile, for those skilled in the art, the specific implementation modes and application ranges of the embodiments of the application will be changed, and the content of the description should not be understood as a limitation of the application.

Claims

1. A railway tunnel structure point cloud decoupling separation method, characterized in that, The method comprises the following steps: Step S1, voxelization downsampling processing is performed on the original railway tunnel point cloud to obtain a single-layer tunnel point cloud model; Step S2, cloth simulation filtering algorithm is applied to separate the point cloud of the tunnel bottom track area; Step S3, the remaining tunnel point cloud is divided into three independent structural blocks, i.e., a left sidewall block, a right sidewall block and a vault block; Step S4, directional adjustment is performed on each structural block, the best fitting plane of each point cloud block is calculated through principal component analysis, a rotation matrix is calculated based on the normal vector of the plane, and the main plane of each point cloud block is rotated to be parallel to the horizontal reference surface; Step S5, cloth simulation filtering algorithm is applied to each point cloud block after rotation transformation to separate the tunnel sidewall structure point cloud and the non-structure object point cloud attached thereto; Step S6, the inverse matrix of the corresponding rotation matrix is applied to each tunnel sidewall point cloud block extracted to restore it to the original coordinate system; Step S7, the three tunnel structure point cloud blocks after inverse transformation are spliced to form a complete tunnel main structure point cloud, and uniformization processing is performed to obtain a fine separated tunnel structure surface point cloud; In step S3, dividing the remaining tunnel point cloud into three independent structural blocks comprises the following sub-steps: Step S31, a copy of the original point cloud is created, the copy point cloud is vertically cut along the tunnel axis direction to divide it into left and right parts, which are respectively used as the point cloud of the left sidewall block and the right sidewall block; Step S32, the original point cloud which has not been divided is vertically cut along the tunnel axis direction and perpendicular to the division line in step S31 to divide it into upper and lower parts, and the upper part point cloud is reserved as the point cloud of the vault block; In step S4, the directional adjustment of each structural block further comprises calculating the center of the minimum bounding box of each point cloud block and taking it as the rotation center of the rotation transformation.

2. The railway tunnel structure point cloud decoupling separation method according to claim 1, characterized in that: In step S1, in the voxelization downsampling processing, the voxel size is set to be higher than the nominal accuracy of the point cloud collection instrument.

3. The railway tunnel structure point cloud decoupling separation method according to claim 1, characterized in that: In step S2, when the cloth simulation filtering algorithm is applied to separate the point cloud of the tunnel bottom track area, the grid size is set to be less than or equal to 0.1 meters, the separation threshold is set to be 0.2 meters to 0.5 meters, and the separation threshold is strictly higher than the actual paving surface height of the track.

4. The railway tunnel structure point cloud decoupling separation method according to claim 1, characterized in that: In step S5, when the cloth simulation filtering algorithm is applied to each point cloud block after rotation transformation, the grid size is set to be less than or equal to 0.1 meters, and the separation threshold is set to be less than or equal to 0.1 meters.

5. The railway tunnel structure point cloud decoupling separation method according to claim 1, wherein: In step S5, the non-structure object point cloud comprises a catenary, a cable pipeline and a distribution box.

6. The railway tunnel structure point cloud decoupling separation method according to claim 1, characterized in that: In step S6, the uniformization processing is performed using the initial voxel size. 7.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is configured to perform the method according to any one of claims 1-6 when the computer program is executed by the processor. The processor executes the computer program to implement the railway tunnel structure point cloud decoupling separation method in any one of claims 1 to 6.

8. A computer readable storage medium storing a computer program, characterized in that, The computer program is executed by the processor to implement the railway tunnel structure point cloud decoupling separation method in any one of claims 1 to 6.

Images